Missed lesions represent a significant source of error in film reading. A perceptual model of the film-reading process in radiology is proposed which identifies three major causes of omission errors: (1) failure to apply appropriate decision-making criteria to searched and recognized targets. A series of experiments are proposed to test the model and determine what proportion of errors fall into each error category. The film-reading process will be studied by monitoring eye movements of radiologists as they search chest films for nodules in order to measure the effectiveness of visual search and target recognition. ROC analysis will be used to quantify decision accuracy. A computer controlled liquid crystal viewscreen will be used to highlight selected areas of the chest film. The effectiveness of this type of visual feedback in reducing search and recognition errors will be evaluated. The results of these experiments will clarify the role of perceptual processes in film reading and suggest ways of reducing errors thus improving diagnostic accuracy.